Near-maximum-likelihood detection in spatially correlated MIMO systems using semi-definite relaxation with cutting planes

Abstract

In this paper, a new efficient scheme is proposed for the fast detection in spatially correlated MEMO systems. By adding cutting planes to the original semi definite relaxation scheme, the new method shows more robustness than other fast detection schemes towards the increasing condition number of the channel matrix, which is always caused by the correlation between different antennas. In order to prove its advancement, the bit error rate performance of the new scheme is compared with other detection schemes like lattice reduction, serial interference cancelation and the original semi definite relaxation. Simulation results show clearly that with the ill conditioned channel matrix, the new detection strategy outperforms the other schemes, while its complexity remains cubic to the number of antennas plus the number of cutting planes added.